Jazz Genius

The Negative Sentiment of Mamie Smith

Jeremy L. McLaughlin

Our dataset only includes three songs from Mamie Smith and four from Marion Harris, but the sentiment analysis for these two artists was markedly different than the rest of the performers in our dataset. Their negative sentiment scores mark them as just two of three artists with a negative average score (as noted here, Bill “Bojangles” Robinson is the other, with the lowest score based on just one song).

What might account for this negative sentiment of these two artists’ lyrics? The musical proclivity of Mamie Smith provides an interesting point of initial analysis and discussion for this question.

Mamie Smith was a multi-faceted performer who excelled in various styles of dance and music. She was the first African American artist to make a blues record in 1920 that stuck to a strictly Afro-American standard of music (compared to other African Americans who had performed European-centric music for Euro-American audiences). This paved the way for other blues singer who straddled the lines of musical genres and helped to usher in the era of classical female blues in the 1920s and 1930s.

This close association with the blues may explain why Smith’s lyrics are given a negative sentiment rating by our sentiment analysis model. Our dataset and the songs within it were chosen based on the artists’ name and their association with jazz, but whether the individual songs fall into the category of jazz or another song genre (or more than one) is certainly a point of discussion for the project going forward.

Also, we did not train the data with our lyrics but used the VADER sentiment analysis script, which is based on social media postings. It makes sense that the model would recognize lyrics like this – from Smith’s Don’t Care Blues – as having a negative emotional sentiment:

I haven't eat for a week, I even walk in my sleep
(Bugs buzz around me, fleas always creep?)
Every night and day I hang my head and weep
Yes, and weep
I love my baby, I want him for myself
If he don't be mine, he won't have nobody else
He has left me, I do the best I can
I'm a do-right kid, (get a friend and replace that man?)

Regardless of the training data or script, these lyrics feel “sad” or emotionally negative from a text classification perspective. The sentiment analysis accurately reflects this with a compound sentiment score of -0.8669 – very negative.

Smith’s Lonesome Mama Blues has a similar feel, with lyrics that speak of “taking my man away” and “I must sigh…and cry” and a grieving heart. However, the sentiment analysis missed the mark with that song, giving it an amazingly high compound sentiment score of 0.9198 – very positive.

The analysis mis-classifies Smith’s third song, That da da strain which espouses the benefits and positive feelings we all get from hearing a good song.

Have you heard it, have you heard it
That Da Da Strain?
It will shake you, it will make you
Really go insane
Everybody's full of pep
Makes you watch your every step
Every prancer, every dancer
Starts to lay 'em down

The lyrics of That da da strain are very repetitive but somehow was classified with a compound score of -0.2552 – somewhat negative. Coupled with the very high, and falsely positive sentiment score of Lonesome Mama Blues, Smith ends up with a compound artist sentiment score higher than the one song from Bill “Bojangles” Robinson (but still in the negative overall categorization).

These results certainly provide an initial jumping off point for further analysis of individual artists and their lyrics and the overall effectiveness of the sentiment analysis model. As Ben notes, it is hard to compare an artist with only a handful of songs to one that has 50 or 100 songs in our dataset. The effectiveness of the model and balancing the various songs of an individual artist is an interesting point for follow up as it directly affects the way Mamie Smith is compared to Bill “Bojangles” Robinson and the weighted value of Smith’s three songs in the final sentiment score.